[DO NOT MERGE] Test if updated kubespawner resolves culler failure #3149
Conversation
|
Merging this PR will trigger the following deployment actions. Support and Staging deployments
Production deployments
|
| @@ -27,8 +27,8 @@ jupyterhub: | |||
| url: http://www.ucmerced.edu/ | |||
| singleuser: | |||
| image: | |||
| name: quay.io/2i2c/2i2c-hubs-image | |||
| tag: "14107b8a85fb" | |||
| name: quay.io/2i2c/culler-failure | |||
There was a problem hiding this comment.
This should actually be setting the image under hub.image rather than singleuser.image
There was a problem hiding this comment.
I opened #3150 about having a separate staging hub for ucmerced perhaps, given they're still on the old user image and there may be other things we need to test.
There was a problem hiding this comment.
Woof, is it too late in the game to change that image name? I was definitely lead to believe I was changing the right thing since it was the "hubs-image"
There was a problem hiding this comment.
That name has confused me also, but since we are phasing it out I think we shouldn't work on changing the name at this point.
I wonder if we have this failure-to-cull result associated with using that singleuser image and/or configuring a specific jupyter server application. It could be that the user servers incorrectly reports they are active when they arent, and that is fixed by a modern jupyter server, but we are stuck with old versions in this hub because we use this old image?
|
Hi @sgibson91, I'm an admin at UC Merced, and I think this Test image is causing disruption to our users, which I would like to bring to your attention. https://2i2c.freshdesk.com/support/tickets/980 I was able to reproduce the issue as well. |
|
Hey @NessieCanCode! I've undeployed this change, and things should be working now! I also replied on the ticket :) |
|
@NessieCanCode also, welcome to our github :) |
Investigating culler fialures
sgibson91
force-pushed
the
culler-failure-experimental-image
branch
from
05ea398
to
96f2a27
Compare
|
I reverted this and redeployed from master again |
ref #3130
I have opened this for visibility into what I have done, but shouldn't be merged. If updating kubespawner fixes the issue, then we should aim for a new release and update it in the main hub image, not maintain another image.
We are hoping that with the deployment of this image, some long-lived pods will be successfully culled.